[Intel-gfx] [v2] drm/i915/tgl: Add Wa_1606054188:tgl

[Ville Syrjälä]

On Fri, Jan 31, 2020 at 10:19:51PM -0500, Matt Atwood wrote:
> On Tiger Lake we do not support source keying in the pixel formats P010,
> P012, P016.
> 
> v2: Move WA to end of function. Create helper function for format
> check. Less verbose debugging messaging.
> 
> Bspec: 52890
> Cc: Matt Roper <matthew.d.roper at intel.com>
> Cc: Manasi Navare <manasi.d.navare at intel.com>
> CC: Ville Syrjälä <ville.syrjala at intel.com>

Wrong address

> Signed-off-by: Matt Atwood <matthew.s.atwood at intel.com>
> ---
>  drivers/gpu/drm/i915/display/intel_sprite.c | 22 +++++++++++++++++++++
>  1 file changed, 22 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_sprite.c b/drivers/gpu/drm/i915/display/intel_sprite.c
> index 2f277d1fc6f1..6e4d73588b48 100644
> --- a/drivers/gpu/drm/i915/display/intel_sprite.c
> +++ b/drivers/gpu/drm/i915/display/intel_sprite.c
> @@ -2070,6 +2070,18 @@ vlv_sprite_check(struct intel_crtc_state *crtc_state,
>  	return 0;
>  }
>  
> +static bool intel_format_is_p01x(int format)
                                    ^^^
u32

> +{
> +	switch(format){

Missing spaces

> +		case DRM_FORMAT_P010:
> +		case DRM_FORMAT_P012:
> +		case DRM_FORMAT_P016:
> +			return true;
> +		default:
> +			return false;

Wrong indentantion

> +	}
> +}
> +
>  static int skl_plane_check_fb(const struct intel_crtc_state *crtc_state,
>  			      const struct intel_plane_state *plane_state)
>  {
> @@ -2143,6 +2155,16 @@ static int skl_plane_check_fb(const struct intel_crtc_state *crtc_state,
>  		return -EINVAL;
>  	}
>  
> +	/* Wa_1606054188:tgl
> +	 *
> +	 * TODO: Add format RGB64i when implemented.
> +	 *
> +	 */

Wrong comment format

> +	if(IS_GEN(dev_priv, 12) &&

Missing space

> +	   plane_state->ckey.flags & I915_SET_COLORKEY_SOURCE &&
> +	   intel_format_is_p01x(fb->format->format))
> +		DRM_DEBUG_KMS("Source color keying not supported with P01x formats\n");

Missing error return.

Was also going to suggest you should use drm_dbg_kms() now, but
looks like this file hasn't been converted yet. Well, I guess
there'd be no harm in using drm_dbg_kms() anyway if you want.

As for the w/a itself, not sure it's any more broken than any other
planar format (don't have the hw to test it right now). But I just
tried my wip colorkey test on glk with nv12/p010 and while it
more or less seems to work the chroma upsampling is definitely
making it impossible to test with crcs. I guess we'll get to
testing out eventually to see if it is actually more broken than
that. Certainly wouldn't be the first time the hw has issues with
the >8bpc to 8bpc conversion for the key match. Just a bit
surprising that it would be limited to just the specific combo
of tgl and P01x formats.

> +
>  	return 0;
>  }
>  
> -- 
> 2.21.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Ville Syrjälä
Intel